Objective. The age at which body mass index (BMI) increases after its nadir in childhood, adiposity rebound (AR), is a critical period for the development of obesity. Children with early AR are at substantially increased risk of adult obesity. Few studies have examined the factors that influence the timing of the AR. The aims of this study were to test for influences on the timing of the AR, and to test the hypothesis that early AR is promoted by high-protein intake.
Design and Participants. Longitudinal cohort study of 889 children representative of the United Kingdom, followed from birth to 5 years.
Main Outcome Measures. We tested for differences in dietary intake, parental BMI, socioeconomic status, and childhood BMI between 3 groups of children characterized by the following: very early AR (at or before 43 months), early AR (from 49 but before 61 months), and later AR (after 61 months).
Results. There was no evidence of associations between dietary protein intake, or any other dietary variable, and timing of the AR. Children with very early AR and early AR had parents with significantly higher BMI, and were significantly more likely to have at least 1 obese parent.
Conclusion. This study does not support the hypothesis that early AR is promoted by high-protein intake. None of the dietary variables tested were significantly associated with timing of the AR, and timing of AR was not associated with socioeconomic status. Parental obesity was associated with an earlier AR.
There is now considerable evidence that adiposity rebound (AR), the point at which the body mass index (BMI) increases after its nadir in childhood, is a critical period for the development of obesity.1 An early AR (younger age at onset of AR) is associated with higher BMI in adolescence and young adulthood,2–4 and substantially increased risk of adult obesity, as defined by BMI.5
At present, the factors that influence the timing of the AR are unclear, and research aimed at identifying these is indicated.5 An understanding of which factors influence the timing of AR might facilitate obesity prevention, and might also identify important mechanisms of energy balance regulation. In a small study of French children (n = 112) higher protein intake at age 2 years was associated with earlier AR,6 and it was suggested that the association was causal. The primary aim of the present study was to test this hypothesis using a larger sample of children. A secondary aim was to assess the influence of a number of other factors on timing of the AR in a large-scale prospective cohort study, the Avon Longitudinal Study of Pregnancy and Childhood (ALSPAC7).
ALSPAC is a prospective cohort study of the determinants of health during childhood in the Bristol-Avon area of the United Kingdom (UK).7 The cohort has been described in detail elsewhere7,,8 and is broadly representative of the UK population in terms of socioeconomic status, although with a slight underrepresentation of ethnic minority families, and slight overrepresentation of wealthier families.8 Summary anthropometric measurements of the cohort from infancy and early childhood did not differ significantly from UK reference data,9 and in fact were almost identical to these data.8 Subjects used in the analyses for the present study represented a 10% subsample of the cohort, randomly selected for more detailed investigations. These children (n = 889) were born in 1991 and 1992 and are representative of the entire cohort.7,,8 Measurements on this subsample were made at frequent intervals: birth, 4, 8, 12, 18, 24, 31, 37, 43, 49, and 61 months.
Height, Weight, BMI, and AR
At each measurement occasion up to 24 months, length was measured to 0.1 cm using the Harpenden Neonatometer and Holtain Kiddimetre (Holtain, Dyfed, Wales). From 24 months height was measured to 0.1 cm using the Leicester Height Measure (Cranlea, Birmingham, United Kingdom). Weight was measured to 0.1 kg at each measurement occasion in underwear using SECA scales (SECA Ltd, London, United Kingdom). BMI was then calculated for each measurement period.
Timing of AR was assessed, as a categorical variable, by visual inspection, using the method previously described.2,,3 This involved identifying an upward trend in BMI after its nadir. To identify that such a trend had occurred (and to distinguish it from random fluctuation and measurement error) it was necessary to: 1) specify that all consecutive measurements of BMI after the nadir showed an increase, and 2) require that any increase in BMI after the nadir had to equal or exceed 0.1 kg/m2. In fact, changes in BMI between measurements of such a small magnitude around the time of AR were rare, but this additional criterion was included to reduce any subjectivity in the assessment and to avoid basing judgements on changes in BMI which were within the limits of measurement error. This definition of rebound was therefore slightly more conservative than the definitions in the literature on which it was based.2,,3 We established that our definition was the same as that of the literature method on which it was based by consulting the principal author of the original report2and conferring over the BMI versus age plot for 9 children who comprised a wide range in timing of AR. This procedure confirmed that our methods for assessing the timing of AR produced the same conclusions.
The BMI was expressed as a standard deviation score (SDS), also known as the Z score, relative to contemporary UK reference data9 (Child Growth Foundation Software, London, UK).
Dietary and Nutrient Intake
Dietary intake data were obtained in all children by 3-day prospectively collected household measures records at ages 8 and 18 months. To replicate the earlier study6 presentation of the analysis is confined here to the dietary data at 18 months, although conclusions did not differ when dietary data were analyzed at the other time period. From these records we calculated absolute intake of energy and macronutrients, energy intake per kg body mass, and macronutrient intake as a percentage of dietary energy intake using the most recent computerized version of McCance and Widdowson's The Composition of Foods (Royal Society of Chemistry, Cambridge, UK).
Parental heights (to 1 cm) and weights (to 0.5 kg, pregravid weight of mothers) were obtained by self-report.7Socioeconomic status (social class category) was determined in the standard way10 from the occupation of the mother's partner. Educational attainment of the mother was obtained by self-report and summarized in 5 categories from lowest (category A) to highest (category E). These 5 categories were as follows: A, Statutory school attendance to age 16 years, with no formal qualifications attained; B, Attainment of technical or vocational qualifications by age 16 years; C, Attainment of at least 1 pass in academic examinations taken at age 16; D, Attainment of academic examinations taken at age 18; and E, Attainment of university degree.
The aim of the analysis was to test for differences between children with earlier AR and the rest, using timing of the AR as a categorical variable. Very early AR was defined here as at or before 43 months; early AR from 49 but before 61 months; and later AR after 61 months. Although these definitions were somewhat arbitrary, we confirmed that our results were not a function of the precise categories chosen by repeating the analyses with different combinations of categories to represent different timing of AR. Differences between categories distinguished based on continuous variables were tested for significance using 1-way analysis of variance (ANOVA) with follow-up multiple comparisons incorporating Bonferroni corrections as appropriate. Differences for categorical variables were tested for significance using χ2 tests.
Dietary and Nutrient Intake at 18 Months of Age
Dietary and nutrient intakes at age 18 months, available for 772 of the 889 subjects, are given in relation to timing of the AR in Tables 1 (absolute intakes) and 2 (relative intakes). There was no evidence of differences in dietary intake associated with timing of the AR. There was no evidence of any association between protein intake and timing of the AR.
There was no evidence of differences in absolute BMI between those with very early or early AR and the rest at any age before 4 years (ie, before the rebound). There was some evidence that girls with very early rebound had higher BMI SDS before the rebound than the rest. After the rebound had occurred BMI SDS and absolute BMI were significantly higher in those children who had rebounded very early (1-way ANOVA;P < .001; Fig 1).
Mean BMI of parents was significantly higher in those children with very early rebound (Table 3). Having at least 1 obese parent (defined as BMI ≥30.0) was also significantly associated with both very early and early AR.
Parental Education and Socioeconomic Status
There was no evidence of any associations between parental educational attainment or socioeconomic status, and timing of the AR (Table 3).
There is currently considerable interest in critical periods for obesity development1 because these might offer opportunities for obesity prevention, and a way of understanding the mechanisms by which energy balance is regulated. Although the AR constitutes only 1 of 3 such critical periods,1 the timing of the AR does have a significant influence on BMI in adolescence and adulthood, and risk of adult obesity as defined by BMI.2–5 Rolland-Cachera et al6 found that an early AR was associated with high-protein intake, and speculated that this relationship was causal.11 This was based on a relatively small study (112 subjects) and the authors concluded that their findings should be confirmed by other studies. To our knowledge, the present study is the only attempt to replicate that of Rolland-Cachera et al.6 The present study was a large-scale, prospective test of the hypothesis that higher protein intake promotes earlier AR. We found no evidence to support this hypothesis, although group mean values for the percentage of energy from protein (Table 2) were not quite as high as those of the children who experienced early AR in the Rolland-Cachera study.6 In fact, there was no clear evidence that any of the dietary variables were related to timing of the AR (Tables 1 and 2). This result, like the others reported in the present study, was independent of the categories chosen to represent differences in timing of the AR: our conclusions were not altered when different combinations of categories were chosen to represent differences in timing of the AR.
A number of possible mechanisms of early AR have been proposed. These include parental control over feeding in early childhood, which can be counterproductive and lead to impaired ability to regulate energy intake.12 Although we did not test this hypothesis directly, there was no evidence of any association between energy intake and timing of the AR in this study. It has been suggested that early AR might be associated with gestational diabetes,13and might be a marker for early maturation1,,14 but we did not measure these variables. Our evidence of no association between socioeconomic status and timing of the AR is consistent with one recent European study15 and is potentially important given current concerns over the interrelationships between socioeconomic status and obesity.
There were significant associations between parental BMI, parental obesity (defined as BMI ≥30.0), and earlier AR (Table 3), and, indeed, these were the only variables we tested that were positively and significantly associated with earlier AR. This observation provides further emphasis that the child of obese parents could be a useful target for obesity prevention efforts. However, it should be noted that parental BMI was less confidently measured than the other variables because it was based on self-report. Self-reports of height, and to a greater extent weight, are subject to small but significant biases.16 It is also worth noting that the effect of timing of AR on risk of adult obesity is independent of parental obesity.5
The present study was considerably larger than previous investigations of the causes/correlates of early AR. Children who had very early AR in our study were present in relatively large numbers, and they experienced AR around 2 years earlier than mean values from studies in the literature (mean values of AR typically 5–7 years;2–5). This might, in part, reflect a secular trend to earlier AR9: previous studies were conducted on cohorts that experienced AR >20 to 30 years ago;2–5 our cohort experienced AR in the mid-1990s and AR occurs earlier now than in the past.9 Therefore, it seems likely that children in our category of very early AR were unusual at least in the sense that their AR occurred early relative to that of their peers, but even in this group there was little evidence of differences in the variables measured, with the exception of parental obesity. It is conceivable that genetic factors might be important in influencing timing of the AR,1 but the relative importance of genotype and environment on the AR remain unclear.5 The secular trend to earlier AR9 suggests that there are strong environmental influences on the timing of the AR.
In summary, the present study suggests that dietary energy and macronutrient intakes are not associated with timing of the AR. In particular, there was no evidence that higher protein intake was associated with early AR. Timing of AR was not associated with socioeconomic status. The only factor significantly associated with timing of the AR in the present study was parental BMI. Children of parents with high BMI, or with at least 1 obese parent, were significantly more likely to have an earlier AR. Parental obesity is a well-established risk factor for obesity17 and the present study suggests that it may operate, at least in part, via an influence on the timing of the AR. In the absence of any other markers of early AR, the factors currently known to be associated with obesity risk, including age and parental obesity,16 inactivity/watching television,18 and preschool activity19 should continue to be used as the principal targets of obesity prevention measures in children.
ALSPAC was supported by the Wellcome Trust, Department of Health, Ministry of Agriculture Fisheries and Food, Medical Research Council, British Gas, and other companies.
Ahmad Dorosty was supported by the Iranian Ministry of Health and Medical Education.
We thank all of the parents and children who took part in the study, and the ALSPAC Study Team, which consists of interviewers, technicians, clerical workers, scientists, volunteers, and managers.
In addition, we thank Dr Marie Francoise Rolland-Cachera for her help and advice on the methodology for defining timing of the AR. We also thank Dr Jan Love of the Robertson Center for Biostatistics for statistical advice, and the Child Growth Foundation, London, for providing the growth monitoring equipment.
- Received April 6, 1999.
- Accepted September 8, 1999.
Reprint requests to (J.J.R.) Department of Human Nutrition, Yorkhill Hospitals, Glasgow G3 8SJ, Scotland. E-mail:
- AR =
- adiposity rebound •
- BMI =
- body mass index •
- ALSPAC =
- Avon Longitudinal Study of Pregnancy and Childhood •
- UK =
- United Kingdom •
- SDS =
- standard deviation score •
- ANOVA =
- analysis of variance
- Rolland-Cachera MF,
- Sempe M,
- Guillard-Bataille M
- Prokopec M,
- Bellisle F
- ↵Whitaker RC, Pepe MS, Wright JA, et al. Early adiposity rebound and risk of adult obesity. Pediatrics. 1998;101(3). URL: http://www.pediatrics.org/cgi/content/full/101/3/e5
- Golding J,
- and the ALSPAC Study Team
- ↵ALSPAC. Children of the nineties. Available at: http://alspac2.ich.bris.ac.uk/ALSPACext
- Cole TJ,
- Freeman JV,
- Preece MA
- ↵Office of Population Censuses and Surveys. Standard Occupation Classification. 1st ed. London, United Kingdom: Her Majesty's Stationery Office; 1990
- Johnson SL,
- Birch LL
- De Spiegelaare M,
- Dramaix M,
- Hernart P
- Rowland ML
- Gortmaker SL,
- Must A,
- Sobol AM,
- et al.
- Moore LL,
- Uyen-Sa DTN,
- Rothman KJ,
- Cupples LA,
- Ellison RC
- Copyright © 2000 American Academy of Pediatrics